DeepSeek Harness PluginManifest Valid

dsh-pseudo-vision

Local OCR, color statistics, pixel scan, and image metadata bridge for text-only DeepSeek Harness providers.

VisionDeveloper Tools

2GitHub Stars0ForksUpdated2026-09-04

Install

$ dsh plugin --profile web add github:DDDFXYqiming/dsh-pseudo-vision

Plugin Overview & Capabilities

AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.

source-grounded

dsh-pseudo-vision provides a local pseudo-vision layer for text-only LLM providers in DeepSeek Harness. When image attachments are sent to non-multimodal models, it intercepts the request, processes the image using local Tesseract OCR and Sharp pipelines (extracting text, layout, color stats, pixel focus, and metadata), injects this structured context, and forwards the text to the provider without calling external vision APIs.

Key Capabilities

Generates pseudo-vision sibling routes for configured text-only LLM providers
Extracts text with normalized bounding box coordinates via an optimized Tesseract.js pipeline
Analyzes color distributions, average luminance, and pixel focus histograms locally using Sharp
Provides low-confidence OCR re-scans, CJK space collapsing, and number verification
Passes through native vision requests transparently while converting text-only image requests into structured textual context

Useful For

Enabling image and screenshot understanding on text-only LLM endpoints within DeepSeek Harness
Extracting logs, code snippets, settings menus, and tabular data from local screenshots without third-party vision APIs
Inspecting image structure, color balance, and layout metadata directly in prompt dispatch pipelines

Who It Fits

DeepSeek Harness users relying on text-only LLM providers who need local image understanding
Developers seeking zero-cloud, privacy-preserving multimodal preprocessing for code and screenshot inspection

Documented Limitations

Cannot replace full visual semantic reasoning of genuine multimodal foundation models
Requires local Node.js environment with native module support for Sharp and Tesseract.js
Enabling bridgeOtherProviders creates duplicate entries in the model selector

DSH Compatibility

Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.

Not tested yetNo runtime compatibility tests have been published yet.

Security Signals

Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.

No automated security signals have been published yet.

Source & Registry Notes

Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.

Source repository: DDDFXYqiming/dsh-pseudo-vision
Registry source: GitHub · dsh-plugin topic
Registry classification: Plugin
Source checked: 45d0bc7 · 2026-09-12

This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.